EP1325415B1

EP1325415B1 - Recovery of file system data in file servers mirrored file system volumes

Info

Publication number: EP1325415B1
Application number: EP01979574A
Authority: EP
Inventors: Srinivasan Viswanathan; Steven R. Kleiman
Original assignee: Network Appliance Inc
Current assignee: NetApp Inc
Priority date: 2000-10-04
Filing date: 2001-10-04
Publication date: 2005-08-03
Anticipated expiration: 2021-10-04
Also published as: WO2002029572A3; US20040153736A1; DE60112462D1; DE60112462T2; US20020194529A1; WO2002029572A9; US7096379B2; US6654912B1; WO2002029572B1; WO2002029572A2; EP1325415A2; WO2002029572A8

Abstract

The invention provides a method and system for recovery of file system data in file servers having mirrored file system volumes. The invention makes use of a 'snapshot' feature of a robust file system (the 'WAFL File System) to rapidly determine which of two or more mirrored volumes is most up-to-date, and which file blocks of the most recent mirrored volume have been changed from each one of the mirrored file systems. In a preferred embodiment, among a plurality of mirrored volumes, the invention rapidly determines which is the most up-to-date by examining a consistency point number maintained by the WAFL File System at each mirrored volume. The invention rapidly pairwise determines what blocks are shared between that most up-to-date mirrored volume and each other mirrored volume, in response to a snapshot of the file system maintained at each mirrored volume and are stored in common pairwise between each mirrored volume and the most up-to-date mirrored volume. The invention re synchronizes only those blocks that have been changed between the common snapshot and the most up-to-date snapshot.

Description

Background of the Invention

1. Field of the Invention

The invention relates to recovery of file system data in file servers having mirrored file system volumes.

2. Related Art

Network file servers and other file systems are subject to errors and other failures, including those arising from hardware failure, software error, or erroneous configuration. Because of the possibility of error, many file systems provide additional copies of data in the file system, such as by providing a mirrored file system volume. In a mirrored file system, a first volume provides a first copy of the file system, while a second volume provides a synchronous, second copy of the file system. Thus, if data on the first volume is corrupted or otherwise lost, data from the second volume can be used in its place transparently.
One problem in the known art is that the first volume and second volume of the file system can fail to remain in synchronization. Thus, each volume of the mirrored file system would include a set of files or other objects from a different timestamp (or checkpoint) in the file system history. As a result, the first volume and second volume will no longer serve as accurate mirrors for each other because one is out-of-date. An aspect of this problem is that, after system crashes, it is unknown which of the first volume and second volume is the most recent. Accordingly, it would be desirable to assure that the first volume and second volume of the file system remain synchronized after system crashes. If it is not possible for the first volume and second volume to remain synchronized, it is desirable to rapidly determine which is the most recent version and use efficiently, so as to cause resynchronization.
A first known method is to resynchronize the two mirror copies after system crashes by copying every block. While this method can generally achieve the result of assuring that the first copy and second copy of the file system are synchronized after system crashes, it has the severe drawback that it is very inefficient, as each file block of at least one of the mirror file systems must be copied to the other one of the mirror file systems. When the file system is particularly large, such as one that approaches or exceeds a terabyte in size, this drawback makes this known method untenable due to its incredible latency (and liability to other failures).
WO 00/07104 discloses a method and system for duplicating all or part of a file system while maintaining consistent copies of the file system. The file server maintains a set of snapshots, each indicating a set of storage blocks making up a consistent copy of the file system as it was at a known time. In a preferred embodiment of this disclosure, the snapshots can be manipulated to identify sets of storage blocks in the file system for incremental backup or copying.
A second known method is to maintain a log of regions or file blocks in each mirrored volume that have been changed (sometimes known as "dirty" file blocks). When such a log is maintained, it is only necessary to copy those file blocks that are dirty, rather than an entire mirrored volume. While this method can generally achieve the result otherwise achieved by the first known method, is still subject to at least two drawbacks. First, this method is more complex, in that it requires careful maintenance so as to ensure that the log remains synchronous. Second, the log itself must generally be mirrored for reliability, which of course reintroduces the entire problem or recovery of mirrored files after system crashes. Third, maintaining this additional log increases the latency of every operation. Moreover, such a technique can introduce additional errors in the event that the log is unreliable.
Accordingly, it would be desirable to provide a technique for recovery of file system date in file servers having mirrored file system volumes that is not subject to drawbacks of the known art.

Summary of the Invention

The invention provides a method and system for recovery of file system data in file servers having mirrored file system volumes. In a preferred embodiment, the invention makes use of a consistency point model including a snapshot feature of a robust file system (the "WAFL File System") to rapidly determine which of two or more mirrored volumes is most up-to-date, and which blocks of the most recent mirrored volume have been changed from each one of the mirrored file systems. Among a plurality of two or more mirrored volumes, the invention rapidly determines which is the most up-to-date by examining a most recent consistency point number maintained by the WAFL File System at each mirrored volume. The invention rapidly and reliably determines what blocks are shared between that most up-to-date mirrored volume and each other mirrored volume, in response to a snapshot of the file system maintained at each mirrored volume and are stored in common pairwise between each mirrored volume and the most up-to-date mirrored volume. The invention copies only those blocks that have been changed between the common snapshot and the most up-to-date snapshot. This rapid and reliable comparison of blocks, followed by the efficient transfer of those blocks that have been changed, does not present drawbacks of the known art.
The invention provides an enabling technology for a wide variety of applications for file system recovery using redundant file systems, so as to obtain substantial advantages and capabilities that are novel and non-obvious in view of the known art. Examples described below primarily relate to mirrored file system volumes in a network file server, but the invention is broadly applicable to many different types of redundant file systems, such as those used in RAID subsystems and parallel storage systems.
These objects are solved in an advantageous manner basically by applying the features laid down in the independent claims. Further enhancements being provided by the subclaims.

Brief Description of the Drawings

Figure 1 shows a block diagram of a system for recovery of file system data in file servers having mirrored file system volumes.
Figure 2 shows a process flow diagram of a method for operating a system as in figure 1.

Detailed Description of the Preferred Embodiment

In the following description, a preferred embodiment of the invention is described with regard to preferred process steps and data structures. Embodiments of the invention can be implemented using general-purpose processors or special purpose processors operating under program control, or other circuits, adapted to particular process steps and data structures described herein. Implementation of the process steps and data structures described herein would not require undue experimentation or further invention.

Lexicography

The following terms refer or relate to aspects of the invention as described below. The descriptions of general meanings of these terms are not intended to be limiting, only illustrative.

block ― in general, any collection of data for data objects in a file system.
consistency point ― in general, any point at which the consistency of a file system is assured or recorded.
file server ― in general, any device which responds to messages requesting file system operations.
file system ― in general, any organization or structure of information for storage or retrieval.
file system data ― in general, any information recorded in a file system or an object in a file system.
file system volume ― in general, any mass storage device, or collection thereof, for storage or retrieval of file system objects.
mirrored volume ― in general, any file system volume having a copy of at least a portion of another file system volume.
parallel storage system ― in general, any file system in which data is recorded, in whole or in part, in multiple locations or multiple ways.
RAID subsystem ― in general, any system including a redundant array of mass storage drives.
recovery of file system data ― in general, any recopying or regeneration of information from one memory or storage medium to another.
redundant file system ― in general, any file system in which data is recorded, in whole or in part, with additional information allowing the recovery of at least a portion of that data.
re-synchronize ― in general, any operation in which objects in a file system are reorganized or rewritten to assure that file system objects maintain or restore synchronization.
shared file block ― in general, any file block whose data contents are located on more than one file system volume.
snapshot ― in general, any consistent file system available, in whole or in part, for later retrieval even if the snapshot is not a current consistent file system.
up-to-date ― in general, a measure of recentness of a file system, file system object, or snapshot.
WAFL File System ― in general, a robust file system, or any file system in which at least one snapshot is maintained in addition to a current consistent file system.

As noted above, these descriptions of general meanings of these terms are not intended to be limiting, only illustrative. Other and further applications of the invention, including extensions of these terms and concepts, would be clear to those of ordinary skill in the art after perusing this application. These other and further applications are part of the scope of the invention, and would be clear to those of ordinary skill in the art, without further invention or undue experimentation.

System Elements

Figure 1 shows a block diagram of a system for recovery of file system data in file servers having mirrored file system volumes.
A system 100 includes a file server (or other device) 110, a communication network 120, and a network interface 130. The file server 110 includes a plurality of mirrored file system volumes 111, each of which includes mass storage for recording and retrieving data. Each file system volume 111 includes at least one snapshot 112 according to the WAFL File System. Each snapshot 112 includes a file system information block 113, including a pointer to an entire consistent file system and a consistency point value 114 indicating a sequence in which that snapshot 112 was generated.
Each file system volume 111 also includes an active file system 115, itself associated with a consistent point value 114. In a preferred embodiment, snapshots 112 are made periodically in response to (and as copies of) an active file system 115. Thus, while every snapshot 112 includes a consistent point value 114 from its associated active file system 115, not every active file system 115 is made into a snapshot, and thus not every consistency point value 114 is associated with a snapshot 112.
The file server 110 receives messages 116 requesting to write data or otherwise alter data from the communication network 120 using the network interface 130. In normal operation, the file server 110 parses those messages 116 and writes the same data to both of the active file systems 115 of the mirrored file system volumes 111, so that each of the mirrored file system volumes 111 includes the same active file systems 115, the same snapshots 112, therefore the same data. However, in the event of a system crash or other error, it might occur that one or more of the mirrored file system volumes 111 fails to remain in synchronization with the others, either because its active file system 115 is not up-to-date or its snapshots 112 are not up-to-date.
If one or more of the mirrored file system volumes 111 is not in synchronization with the others, there will be at least one mirrored file system volume 111 having an active file system 115 with a consistency point value 114 larger than all others. This indicates that the associated active file system 115 and the associated file system volume 111 (with the highest consistency point value 114) is the most up-to-date file system volume 111 of all of the mirrored file system volumes 111.
Similarly, for any pair of mirrored file system volumes 111, there will be at least one common snapshot 112 present for them both, thus having the same consistency point value 114 for the common snapshot 112 at each of the two mirrored file system volumes 111. For any pair of mirrored file system volumes 111 A and B, the difference between the common snapshot 112 and the most up-to-date active file system 115 (say, at mirrored file system volume 111 A) can be easily and rapidly determined using the WAFL File System. The file blocks indicated by that difference are the only file blocks necessary for re-synchronization between the pair of mirrored file system volumes 111 A and B.
While each pair (A and B) of mirrored file system volumes 111 will have at least one common snapshot 112, of which one can be compared with the most up-to-date active file system 115, there is no particular requirement that each other pair (A and C, or A and D) of mirrored file system volumes 111 will have the same common snapshot 112 as the first such pair (A and B). However, for each such other pair (A and C, or A and D) of mirrored file system volumes 111, the difference between the common snapshot 112 and the most up-to-date active file system 115 can still be easily and rapidly determined using the WAFL File System; the file blocks indicated by that difference are the only file blocks necessary for re-synchronization between the other pair (A and C, or A and D) of mirrored file system volumes 111.

Method of Operation

Figure 2 shows a process flow diagram of a method for operating a system as in figure 1.
A method 200 includes a set of flow points and a set of steps. The system 100 performs the method 200. Although the method 200 is described serially, the steps of the method 200 can be performed by separate elements in conjunction or in parallel, whether asynchronously, in a pipelined manner, or otherwise. There is no particular requirement that the method 200 be performed in the same order in which this description lists the steps, except where so indicated.
At a flow point 210, the file server 110 is ready to re-synchronize a plurality of mirrored file system volumes 111.
At a step 211, the file server 110 examines the file system information block 113 for each one of the plurality of mirrored file system volumes 111, to determine a single consistency point value 114 which is the maximum for all active file systems 115 at such mirrored file system volumes 111. While it is possible that there will be more than one such mirrored file system volume 111 having an active file system 115 with that maximum consistency point value 114, there is no particular requirement to select one of such mirrored file system volumes 111 in preference to others, as all active file systems 115 with that identical consistency point value 114 will be identical.
At a step 212, the mirrored file system volumes 111 with the maximum consistency point value 114 for an active file system 115 generates a new snapshot 112 for that active file system 115 and having that maximum consistency point value 114. This new snapshot 112 is thus the most up-to-date snapshot 112 and has the maximum consistency point value 114.
At a step 213, for each one of the plurality of mirrored file system volumes 111 (other than the file system volumes 111 with the most up-to-date active file system 115) the file server 110 examines the file system information block 113, to determine a snapshot 112 at that one mirrored file system volume 111 that is common with the mirrored file system volume 111 having the most up-to-date snapshot 112. Thus, the file server 110 determines a closest degree of synchronization between each mirrored file system volume 111 (in turn) and the mirrored file system volume 111 having the most up-to-date snapshot 112.
At a step 214, for each such closest degree of synchronization, the file server 110 determines a difference between the common snapshot 112 and the most up-to-date snapshot 112, thus generating a set of file blocks that have been changed between the common snapshot 112 and the most up-to-date snapshot 112. These changed file blocks are the only file blocks required to be re-synchronized between the common snapshot 112 and the most up-to-date active file system 115.
At a step 215, for each such set of changed file blocks, the file server 110 resynchronizes each mirrored file system volume 111 with the most up-to-date snapshot 112 by copying only the changed file blocks over, thus generating a copy of the most up-to-date snapshot 112 at each mirrored file system volume 111.
In a preferred embodiment, there are only two such mirrored file system volumes 111. The file server 110 needs to make only one comparison to determine the maximum consistency point value 114 for a most up-to-date active file system 115. The file server 110 needs to examine only one pair of mirrored file system volumes 111 for a common snapshot 112. The file server 110 needs to determine only one set of changed blocks between the common snapshot 112 and the most up-to-date snapshot 112. The file server 110 needs to copy only one set of changed blocks from one mirrored file system volume 111 to the other.
However, in alternative embodiments, there may be more than two mirrored file system volumes 111. Those skilled in the art will see, after perusal of this application, that the invention is easily and readily generalized to additional mirrored file system volumes 111, without undue experimentation or further invention.
In a preferred embodiment, the mirrored file system volumes 111 can each be updated to create new active file systems 115 in response to messages 116 requesting file system operations, even while the snapshot 112 at each mirrored file system volumes 111 is being synchronized with the most up-to-date snapshot 112. Thus, the mirrored file system volumes 111 can each perform the full functions of a file server 110 mirrored file system volume 111 even while the re-synchronization is taking place.
After this step, the method 200 has re-synchronized all of the mirrored file system volumes 111 to the most up-to-date active file system 115.
In a preferred embodiment, the method 200 is performed each time the system 100 recovers from a system crash, as part of the crash recovery process. In alternative embodiments, the method 200 may be performed in response to other events, such as in response to a timer, in response to detection of lack of synchronization between the mirrored volumes, or in response to operator command.

Generality of the Invention

The invention has general applicability to various fields of use, not necessarily related to the services described above. For example, these fields of use can include one or more of, or some combination of, the following:

file system recovery using redundant file systems other than mirrored file system volumes
RAID subsystems
parallel storage systems

Other and further applications of the invention in its most general form, will be clear to those skilled in the art after perusal of this application, and are within the scope of the invention. Although preferred embodiments are disclosed herein, many variations are possible which remain within the concept, scope, and spirit of the invention, and these variations would become clear to those skilled in the art after perusal of this application.

Claims

A method including the step of:

examining a plurality of mirrored file system volumes (111) that each include one or more snapshots;
wherein

each said snapshot includes a file system information block (113), including a pointer to an entire consistent file system and a consistency point value (114) indicating a sequence in which that snapshot was generated; and

each said mirrored file system volume (111) includes an active file system (115), itself associated with a consistency point value (114); and further including the steps of:

determining a most up-to-date mirrored file system volume (111) by reference to said consistency point values ;

determining snapshots (112) held in common between said most up-to-date file system volume (111) and at least two of said plurality of mirrored file system volumes (111);

selecting those file blocks changed between said snapshots (112) held in common and said up-to-date file system volume (111); and

re-synchronizing at least two of said plurality of mirrored file system volumes (111) in response to said steps of selecting:
A method as claimed in claim 1, wherein said consistency point value (114) for which said plurality of mirrored file system volumes (111) are examined is a maximal value.
A method as claimed in claim 2, wherein said file system volume (111) associated with said maximal value is used to generate a new snapshot (112).
A method as claimed in claim 3, further including steps of re-synchronizing each of said plurality of mirrored file system volumes (111) with said new snapshot (112).
A method as claimed in claim 2, wherein said maximal value is associated with said most up-to-date file system volume (111).
A method as claimed in claim 1, wherein said step of selecting further includes selecting file blocks responsive to a difference between a snapshot (112) for said most up-to-date file system volume (111) and a common snapshot (112) for said plurality of mirrored file system volumes (111).
A method as claimed in claim 6, further including steps of copying said set of changed file blocks to said plurality of mirrored file system volumes (111).
A method as claimed in claim 7, wherein said set of changed file blocks include a collection of data for data objects in said plurality of mirrored file system volumes (111).
A method as claimed in claim 1, wherein said plurality of mirrored file system volumes (111) include a RAID subsystem.
A method as claimed in claim 1, wherein said plurality of mirrored file system volumes (111) include volumes for parallel stored systems.
An apparatus comprising:

means for examining a plurality of mirrored file system volumes (111) that each include one or more snapshots;
wherein:

each said snapshot includes a file system information block (113), including a pointer to an entire consistent file system and a consistency point value (114) indicating a sequence in which that snapshot was generated; and

each said mirrored file system volume (111) includes an active file system (115), itself associated with a consistency point value (114); and further comprising:

means for determining a most up-to-date mirrored file system volume (111) by reference to said consistency point values;

means for determining snapshots (112) held in common between said most up-to-date file system volume (111) and at least two of said plurality of mirrored file system volumes (111);

means for selecting those file blocks changed between said snapshot (112) held in common and said up-to-date file system volume (111); and

means for re-synchronizing at least two of said plurality of mirrored file system volumes (111) in response to said means for selecting.
Apparatus according to claim 11 including
said plurality of mirrored file system volumes (111), each having at least one snapshot including an entire consistent file system, each said snapshot having a consistency point value (114);
a first comparison element capable of being coupled to a plurality of said consistency point values (114);
a second comparison element, responsive to an output of said first comparison element, said second comparison element being capable of being coupled (a) to a first snapshot associated with said output on a first one of said file system volumes and (b) to a second snapshot associated with a second one of said file system volumes, said second comparison element being capable of providing a selection of file blocks in response thereto.
Apparatus as claimed in claim 12, wherein said second snapshot is held in common between said first one and said second one of said file system volumes.
Apparatus as claimed in claim 12, wherein said output of said first comparison element indicates which snapshot (112) for said plurality of mirrored file system volumes (111) has a maximal consistency point value (114).
Apparatus as claimed in claim 14, further comprising a file server element that generates a new snapshot (112) from a file system volume (111) associated with said maximal consistency point value (114).
Apparatus as claimed in claim 15, wherein said file server element re-synchronizes each of said plurality of mirrored file system volumes (111) with said new snapshot (112).
Apparatus as claimed in claim 14, wherein said maximal consistency point value (114) is associated with a most up-to-date file system volume (111).
Apparatus as claimed in claim 12, wherein said second snapshot is a common snapshot further associated with others of said file system volumes, and wherein said selection of file blocks is responsive to a difference between said first snapshot and said second snapshot.
Apparatus as claimed in claim 18, further comprising a file system element that copies said selection of file blocks to said second one and said others of said file system volumes (111).
Apparatus as claimed in claim 19, wherein said selection of file blocks includes a collection of data for data objects in said file system volumes (111).
Apparatus as claimed in claim 11 or 12, wherein said file system volumes (111) include a RAID subsystem.
Apparatus as claimed in claim 11 or 12, wherein said file system volumes (111) include volumes for parallel stored systems.
A computer program product comprising program code means that, when executed on a computer system, cause the computer system to effect the steps of any one of claims 1 to 10.